Revolution in GPCR signalling: opioid receptor heteromers as novel therapeutic targets: IUPHAR Review 10

Abstract

GPCRs can interact with each other to form homomers or heteromers. Homomers involve interactions with the same receptor type while heteromers involve interactions between two different GPCRs. These receptor–receptor interactions modulate not only the binding but also the signalling and trafficking properties of individual receptors. Opioid receptor heteromerization has been extensively investigated with the objective of identifying novel therapeutic targets that are as potent as morphine but without the side effects associated with chronic morphine use. In this context, studies have described heteromerization between the different types of opioid receptors and between opioid receptors and a wide range of GPCRs including adrenoceptors, cannabinoid, 5-HT, metabotropic glutamate and sensory neuron-specific receptors. Recent advances in the field involving the generation of heteromer-specific reagents (antibodies or ligands) or of membrane-permeable peptides that disrupt the heteromer interaction are helping to elucidate the physiological role of opioid receptor heteromers and the contribution of the partner receptor to the side effects associated with opioid use. For example, studies using membrane-permeable peptides targeting the heteromer interface have implicated μ and δ receptor heteromers in the development of tolerance to morphine, and heteromers of μ and gastrin-releasing peptide receptors in morphine-induced itch. In addition, a number of ligands that selectively target opioid receptor heteromers exhibit potent antinociception with a decrease in the side effects commonly associated with morphine use. In this review, we summarize the latest findings regarding the biological and functional characteristics of opioid receptor heteromers both in vitro and in vivo.

Links to online information in the IUPHAR/BPS Guide to PHARMACOLOGY

TARGETS	LIGANDS
5-HT1A receptor	5′-guanidinonaltrindole, GNTI
α2A-adrenoceptor	5-HT
Apelin receptor, APJ	CCL4 (MIP-1β)
β2-adrenoceptor	CCL5 (RANTES)
Cannabinoid CB1 receptor	CGRP
Chemokine CCR5 receptor	Clonidine
Chemokine CXCR4 receptor	DAMGO
δ receptor, DOP	DPDPE
Dopamine D1 receptor	DSLET
ERK1/2	dynorphin A
Gastrin-releasing peptide receptor (BB2 receptor)	L-779,976
κ receptor, KOP	[Leu]enkephalin
MAPK	LPS
mGlu5 receptor	MPEP
μ receptor	M-MPEP
NK1 receptor	Naltriben
NOP receptor	Nor-binaltorphimine, nor-BNI
protein kinase C (PKC)	Substance P
SNSR4 (MRGPRX1)	U 50488
phospholipase C (PLC)	U 69593
Somatostatin sst2 receptor

This Table lists protein targets and ligands which are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Pawson et al., 2014) and the Concise Guide to PHARMACOLOGY 2013/14 (Alexander et al., 2013a, b).

Introduction

It is becoming generally accepted that GPCRs including opioid receptors interact with each other to form homomers and heteromers, opening new therapeutic possibilities for identifying drugs targeting GPCRs.

The physical interaction between opioid receptors was first reported in 1997 using δ receptors, and these studies showed that agonist treatment modulated the level of receptor homodimers (Cvejic and Devi, 1997). Similar studies also showed that the κ receptors existed as detergent-insensitive dimers (Jordan and Devi, 1999) and μ receptors existed in interacting complexes with distinct trafficking properties (He et al., 2002). The most recent X-ray crystallization analysis of μ receptors (Manglik et al., 2012) has revealed a twofold symmetrical dimer through a four-helix bundle motif formed by transmembrane segments 5 and 6. Although this dimeric arrangement of μ receptors could be due to the conditions used to crystallize the receptor, these structural findings are exciting as they enable the development of structure-based approaches to complement the more conventional drug discovery programmes, in addition to providing novel insights into the roles of oligomerization in GPCR function.

Over the last decade, an increasing number of studies have explored the ability of GPCRs including opioid receptors to heteromerize (either with members of the same family or related families). In the case of opioid receptors, early ‘indirect’ evidence for the presence of heteromers was provided by radioligand binding and electrophysiological studies that suggested interactions between μ and δ receptors (Zieglgänsberger et al., 1982; Rothman et al., 1985; 1988; Metcalf et al., 2012; Akgun et al., 2013). Moreover, studies examining the effect of pretreatment with leucine-enkephalin on morphine-mediated analgesia (Vaught and Takemori, 1979a,b), and of δ receptor antagonists on the development of morphine tolerance and dependence (Abdelhamid and Takemori, 1991; Zhu et al., 1999) further supported a functional interaction between μ and δ receptors. Recent studies using heteromer-selective reagents, such as antibodies (Gupta et al., 2010), ligands (Daniels et al., 2005; Waldhoer et al., 2005; Gomes et al., 2013b) or agents that disrupt the heteromer in vivo (He et al., 2011), have begun to provide ‘direct’ evidence for opioid receptor heteromerization.

In addition to heteromers between opioid receptor types, heteromers involving opioid receptors and other GPCRs including adrenoceptors, metabotropic glutamate receptors, sensory neuron-specific receptors, have been described (Gomes et al., 2013a). In this review, we describe in vitro and in vivo evidence for heteromers involving opioid receptors, the development of heteromer-selective ligands and the therapeutic potential of GPCR heteromers as target molecules for novel drug development.

Heteromers between opioid receptor types

δ and μ receptor heteromers

The δ and μ receptor heteromer (δ-μ heteromer) is the most extensively studied opioid receptor heteromer. The ‘direct’ interaction between δ and μ receptors was first demonstrated by co-immunoprecipitation studies (George et al., 2000; Gomes et al., 2000) where Flag-tagged μ receptors were found to form interacting complexes (∼150 kDa) with myc-tagged δ receptors in HEK-293 cells coexpressing both receptors (Gomes et al., 2000). Endogenous δ-μ complexes could also be detected by co-immunoprecipitation studies with the spinal cord membranes from wild-type, but not δ receptor knockout mice (Gomes et al., 2004). Furthermore, BRET assays demonstrated that both receptors exist within 100Å of each other in live cells, which is close enough to allow for direct receptor–receptor interactions (Gomes et al., 2004) (Table 1).

Table 1.

Heteromers between opioid receptor types^a

Heteromer pair	In vitro heteromer properties (binding, signalling, trafficking)	In vivo effects of reagents targeting heteromers	References
δ-μb,c2013	Detection Co-IP, BRET Binding ↓ in affinity for receptor-selective agonists. Allosteric modulation of μ receptor binding by δ receptor ligands and vice versa. Signalling Potentiation of μ receptor signalling by δ receptor ligands and vice versa. Switch from Gαi/o to either Gαz or Gαq signalling. Coupling to and signalling via β-arrestin 2. Trafficking Protomers internalize independently from each other. Some selective agonists can induce heteromer internalization. Methadone internalizes the δ-μ heteromer and targets it for degradation. MDAN-21 immobilizes the heteromer and prevents its endocytosis. Heteromer expression at the cell surface requires RTP4.	δ-μ heteromer selective antibody Antibody detects ↑ in heteromer levels in various brain regions after chronic morphine administration. Bivalent ligand (MDAN21) Antinociception (i.t, i.c.v. and s.c.) with MDAN21 is more potent than morphine without development of tolerance revealed and it is less rewarding than morphine. Biased agonist (CYM51010) Potent antinociception with reduced development of tolerance. TAT-fusion protein (μTM1-TAT) μTM1-TAT disrupts δ-μ heteromers; ↑ in morphine antinociception with ↓ in development of tolerance.	George et al., 2000; Gomes et al., 2000; 2004; 2011; 2013b; Charles et al., 2003; Daniels et al., 2005; Fan et al., 2005; Law et al., 2005; Hasbi et al., 2007; Lenard et al., 2007; Rozenfeld and Devi, 2007; Decaillot et al., 2008; Gupta et al., 2010; He et al., 2011; Milan-Lobo and Whistler, 2011; Yekkirala et al., 2013
δ-κb2013	Detection Co-IP, BRET Binding ↓ in affinity for δ or κ receptor ligands. Binding co-operativity with δ + κ receptor ligands. Signalling ↑ in signalling with a combination of δ + κ receptor agonists. Trafficking ↓ in δ receptor internalization in cells expressing the δ-κ heteromer.	δ-κ heteromer selective antibody Antibody detects δ-κ heteromers in peripheral sensory neurons following thermal allodynia. Antibody enhances anti-allodynic effects of a δ agonist. Bivalent ligand (KDN-21) No antinociception following i.t administration. Heteromer targeting agonist (6′-GNTI)d Antinociception following i.t. but not i.c.v. administration. Antinociception to 6′-GNTI is more potent than that of U50,488.	Jordan and Devi, 1999; Ramsay et al., 2002; Bhushan et al., 2004; Waldhoer et al., 2005; Berg et al., 2012
μ-κb2013	Detection Co-IP, BRET Binding ↓ in affinity for μ receptor-selective agonists. No change in affinity for κ agonists. Signalling Remains to be determined. Trafficking Remains to be determined.	Heteromer targeting agonist (NNTA) Antinociception (i.v., i.t., i.c.v.); NNTA exhibits more potent antinociception than morphine.	Jordan and Devi, 1999; Wang et al., 2005a; Chakrabarti et al., 2010; Yekkirala et al., 2011
μ-NOPb2013	Detection Co-IP Binding ↑ in affinity for μ receptor agonist. Signalling ↓ in potency of μ receptor agonists. Cross-desensitization of μ receptor-mediated signalling by NOP agonist. Trafficking Remains to be determined.	Heteromer targeting ligand (IBNtxA) IBNtxA exhibits more potent antinociceptive effect than morphine determined with reduced side effects.	Pan et al., 2002; Wang et al., 2005b; Majumdar et al., 2011

^aModified from (Gomes et al., 2013a).

^b,cThese heteromers are useful therapeutic targets, based on the known roles of one or both protomers. Therapeutic targets: ^bAntinociception; ^cAntinociceptive tolerance.

^dA recent report shows that 6′-GNTI exhibits biased agonistic properties at κ opioid receptors (Rives et al., 2012).

Examination of the ligand-binding properties of the δ-μ heteromers showed that the binding affinity of agonists to individual protomers was decreased in δ-μ heteromers when compared with individual receptors (George et al., 2000) (Table 1). Interestingly, this was increased in the presence of agonists or antagonists selective for the partner protomer (Gomes et al., 2000; 2004; 2011). Further analysis revealed that this was due to allosteric modulation; the occupancy of one protomer allosterically modulated ligand binding to the partner protomer by affecting the rate of dissociation of the ligand (Gomes et al., 2011). These changes in the pharmacological properties of δ-μ heteromers compared with δ or μ homomers suggested possible differences in signalling between heteromers and homomers.

Comparison of the intracellular signalling between the δ-μ heteromers and δ or μ homomers showed interesting differences. For example, occupancy of one of the protomers in the δ-μ heteromer enhanced signalling mediated via the activation of the partner protomer (Gomes et al., 2000). In addition, while δ or μ homomers are coupled to and signal via Gα_i/o proteins, studies have reported that the δ-μ heteromers could be associated with a Pertussis toxin-insensitive G-protein such as Gz (George et al., 2000; Fan et al., 2005; Hasbi et al., 2007) and/or with β-arrestin 2 (Rozenfeld and Devi, 2007). The latter study showed that activation of a protomer in δ-μ heteromers leads to β-arrestin 2-mediated signalling. This is characterized by the presence of a second phase of ERK activation that is PKC-independent and can be blocked by small interfering RNA to β-arrestin 2 (Rozenfeld and Devi, 2007) (Table 1). Moreover, δ-μ heteromer-mediated β-arrestin 2 signalling leads to changes in the spatiotemporal dynamics of ERK1/2 phosphorylation. For example, in cells expressing the δ-μ heteromer, ERK1/2 which was phosphorylated following treatment with the μ receptor agonist, [D-Ala², N-MePhe⁴, Gly-ol]-enkephalin (DAMGO), was not translocated to the nucleus (as seen with cells expressing only μ receptors). This leads to the phosphorylation of the cytoplasmic and not the nuclear ERK1/2 substrates and ultimately differential activation of transcription factors (Rozenfeld and Devi, 2007). Finally, a study showed that in GH3 cells expressing δ-μ heteromers, the μ receptor agonist, DAMGO, stimulated Ca⁺²-mediated signalling instead of Gα_i/o-mediated signalling (Charles et al., 2003). Taken together, these studies show that heteromerization increases the repertoire of signalling of μ and δ receptors.

It is well known that following continued exposure to agonists, μ or δ receptors are phosphorylated; this leads to the recruitment of β-arrestin, receptor endocytosis to acidic endosomal compartments and termination of Gα_i/o protein-mediated signalling. In the endosomes, the internalized receptors are either dephosphorylated and recycled back to the cell surface to undergo another round of signalling or are targeted to lysosomes for degradation. Very few studies have investigated how heteromerization modulates the endocytosis of the δ-μ heteromer. While one study found that each protomer in the δ-μ heteromer internalized independently from the other protomer, other studies found that the heteromer as a whole could be endocytosed by some selective agonists (DAMGO, deltorphin II, methadone) but not others, such as D-penicillamine(2,5)-enkephalin (DPDPE) or [(D-Ser², Leu⁵]enkephalin-Thr⁶ (Law et al., 2005; Hasbi et al., 2007; Milan-Lobo and Whistler, 2011). Moreover, methadone, a μ receptor agonist that induces homomer internalization and recycling, induced endocytosis of the δ-μ heteromer leading to its degradation and this effect could be blocked by co-treatment with naltriben, a δ receptor antagonist (Milan-Lobo and Whistler, 2011). Another study found that a bivalent ligand selective for the δ-μ heteromer, MDAN-21 (comprising a δ receptor antagonist pharmacophore, DN-20, separated by a 21-atom spacer from the μ receptor agonist pharmacophore, MA-19) (Table 4), did not induce heteromer internalization (Yekkirala et al., 2013). As co-administration of the individual monovalent pharmacophores (DN-20 and MA-19) induced internalization of δ-μ heteromers, it has been suggested that the spacer arm in MDAN-21 by bridging both protomers in the δ-μ heteromer, immobilizes the latter thereby preventing its endocytosis (Table 1). In addition to modulation of receptor internalization and endocytosis, heteromerization appears to modulate the maturation and cell surface expression of the δ-μ heteromer. In cells coexpressing μ and δ receptors, the heteromer is localized to the Golgi apparatus and its cell surface expression requires a chaperone named receptor transport protein 4 (RTP4) (Decaillot et al., 2008). RTP4 protects the receptor heteromer during folding and maturation from ubiquitination and proteasomal degradation (Decaillot et al., 2008). Further studies are needed to examine if RTP4 affects ligand binding and signalling by the δ-μ heteromer.

Table 4.

List of ligands targeting opioid receptor heteromers

Target Heteromer Pair	Ligands	Pharmacophores	Spacer length	References
δ-μ	Bivalent ligand MDAN21	δ antagonist: DN-21 μ agonist: MA-19	21-atom	(Daniels et al., 2005; Gomes et al., 2013b; Harvey et al., 2012)
	Bivalent ligand L2	δ antagonist:ENTI μ agonist: oxymorphone	optimized for the heteromer(19–22-atom) (not exactly mentioned)
	Bivalent ligand L4	δ agonist: DM-SNC80 μ antagonist: naltrexone
	Biased agonist CYM51010	δ-μ agonist: CYM51010	Not applicable (N/A)
δ-κ	Bivalent ligand KDN-21	δ antagonist: naltrindole (NTI) κ antagonist: 5′-guanidinonaltrindole (5′-GNTI)	21-atom	(Bhushan et al., 2004; Waldhoer et al., 2005)
	Heteromer targeting agonist 6′-GNTI*2012	δ-κ agonist: 6′-guanidinonaltrindole (6′-GNTI)	N/A
μ-κ	Heteromer targeting agonist NNTA	μ-κ agonist: N-naphthoyl-β-naltrexamine (NNTA)	N/A	(Yekkirala et al., 2011)
μ1G-NOP	Heteromer targeting ligand IBNtxA	μ1G-NOP agonist: iodobenzoylnaltrexamide (IBNtxA)	N/A	(Majumdar et al., 2011)
μ-CB1R	Bivalent ligand	μ agonist: α-oxymorphamine CB1 antagonist: SR141716	20-atom	(Le Naour et al., 2013)
μ-CCR5	Bivalent ligand Bivalent ligand 1	μ antagonist: naltrexone CCR5 antagonist: maraviroc	21-atom	(Yuan et al., 2012)
μ-mGluR5	Bivalent ligand MMG22	μ agonist: oxymorphone mGluR5 antagonist: m-methoxy-2-methyl-6-(phenylethynyl) pyridine (M-MPEP)	22-atom	(Akgun et al., 2013)
μ-NK1	Bivalent peptides	Opioid agonist: H-Tyr-D-Ala-Gly-Phe NK1 antagonist: Pro-Leu-Trp-O-3,5-Bzl(CF3)2	N/A	(Largent-Milnes et al., 2013; Vardanyan et al., 2011; Yamamoto et al., 2007)
	Bivalent ligands small molecule	μ agonist: fentanyl NK1 antagonist: L732138	N/A
	Multifunctional μ/δ agonist/NK1 antagonist compound TY027	μ-NK1 agonist: H-Tyr-D-Ala-Gly-Phe-Met-Pro-Leu-Trp-NH-3,5Bn(CF3)2 (TY027)	N/A

^*A recent report shows that 6′-GNTI exhibits biased agonistic properties at κ receptors (Rives et al., 2012); N/A, not available

In order to understand the role of δ-μ heteromers in vivo, heteromer-selective reagents (ligands, antibodies) as well as agents that selectively disrupt the heteromer have been developed. Studies with these reagents suggest a possible involvement of δ-μ heteromers in the development of tolerance to morphine. For example, δ-μ heteromer-selective antibodies detect increased heteromer levels in discrete brain regions following chronic morphine administration (Gupta et al., 2010). Moreover, administration of a membrane permeable peptide TAT peptide (YGRKKRRQRRR) fused to the peptide representing transmembrane domain 1 of μ receptors disrupts the δ-μ heteromer and leads to an increase in morphine-mediated antinociception and a decrease in the development of tolerance to morphine (He et al., 2011) (Table 1). These results suggest that the δ-μ heteromer could be a target for the development of antinociceptive therapeutics as potent as morphine but with lesser side effects such as antinociceptive tolerance and dependence. This is supported by studies with MDANs that differ in the length of the spacer arm between pharmacophores and selectively target the δ-μ heteromer. Of these, MDAN-21 was found to exhibit 100 times more potent antinociception and to be less rewarding than morphine (Daniels et al., 2005; Lenard et al., 2007). Moreover, chronic administration of MDAN-21 did not lead to the development of tolerance or dependence (Daniels et al., 2005). In addition to MDANs, bivalent ligands comprising of a high-affinity μ receptor agonist (oxymorphone) joined by a spacer arm to a low affinity δ receptor antagonist (ENTI) or of a high-affinity μ receptor antagonist (naltrexone) joined by a spacer arm to a low affinity δ receptor agonist (DM-SNC80) have been generated (Harvey et al., 2012) (Table 4). However, the antinociceptive effects of these ligands and their side effects have not been evaluated. More recently, a high throughput screening of a small molecule library for a δ-μ heteromer-selective ligand led to the identification of CYM51010 as a biased δ-μ heteromer agonist (Table 4); this study also showed that the compound exhibited potent antinociception with reduced antinociceptive tolerance (Gomes et al., 2013b). The antinociceptive effect of CYM51010 was significantly blocked by δ-μ heteromer-selective antibodies (Gomes et al., 2013b), which would suggest that, in vivo, this heteromer plays an important role in pain regulation (Table 1). Taken together, these studies support the δ-μ heteromer as a novel therapeutic target for pain attenuation with reduced side effects.

δ and κ opioid receptor heteromers

The δ and κ receptor heteromer (δ-κ heteromer) was the first opioid receptor heteromer to be reported. Co-immunoprecipitation studies using κ receptors tagged with a myc epitope and δ receptors tagged with a Flag epitope detected the presence of interacting complexes only in cells coexpressing both receptors (Jordan and Devi, 1999). In addition, BRET assays showed that the two receptors existed in close proximity and could directly interact in live cells (Ramsay et al., 2002) (Table 1).

Examination of the binding, signalling and trafficking properties of the δ-κ heteromer showed that they were distinct from δ or κ homomers. For example, the binding affinities for δ or κ receptor agonists were lower at δ-κ heteromers compared with the respective receptor homomers (Jordan and Devi, 1999). An increase not only in the binding affinity but also in intracellular signalling was observed when the δ-κ heteromer was treated with a combination of δ and κ receptor agonists (Jordan and Devi, 1999) (Table 1). An increase in the binding affinity was also observed when the δ-κ heteromer was treated with a combination of δ or κ receptor antagonists (Jordan and Devi, 1999). These findings suggest the possibility of allosteric interactions between δ and κ receptors. With regard to the trafficking properties of the δ-κ heteromer, studies with etorphine, a potent non-selective opioid agonist that binds to both κ and δ receptors, show that it does not induce δ receptor internalization in cells expressing the δ-κ heteromer, while it induces receptor internalization in cells expressing only δ receptors (Jordan and Devi, 1999). These findings suggest that δ-κ heteromerization alters the trafficking of properties of δ receptors (Table 1).

In order to understand the role of δ-κ heteromers in vivo, heteromer-selective reagents such as antibodies have been generated. A recent study detected the presence of δ-κ heteromers in peripheral sensory neurons following thermal allodynia (Berg et al., 2012). Moreover, the study found that a δ-κ heteromer-selective antibody could enhance the anti-allodynic effects of the δ receptor agonist, DPDPE; this suggests a role for δ-κ heteromers in modulation of thermal allodynia (Berg et al., 2012) (Table 1). Ligands targeting the δ-κ heteromer include a bivalent ligand named KDN-21 and 6′-guanidinonaltrindole (6′-GNTI) (Table 4). KDN-21 comprises a κ receptor-selective antagonist pharmacophore, 5′-guanidinonaltrindole, which is tethered through a 21-atom spacer arm to the δ receptor-selective antagonist pharmacophore, naltrindole. KDN-21 exhibits selective δ1 receptor and κ2 receptor antagonistic activity and does not induce antinociception (Bhushan et al., 2004) (Table 1). In contrast, 6′-GNTI functions as an agonist that induces ∼50-times more potent antinociception than the selective κ receptor agonist U50488H when administered intrathecally but not intracerebroventricularly (Waldhoer et al., 2005) (Table 1). However, a recent study reported that 6′-GNTI also exhibits biased agonistic properties at κ receptors (Rives et al., 2012); this would suggest that the behavioural outcomes from studies with 6′-GNTI could be due to its activity at either κ receptors alone or δ-κ heteromers. Taken together, these studies suggest that δ-κ heteromers form a distinct functional signalling unit that could provide a target for the development of tissue-selective opiate analgesics.

μ and κ receptor heteromers

A few studies have investigated possible heteromerization between μ and κ receptors (μ-κ heteromers). Early co-immunoprecipitation studies using antibodies to the epitope tags on the receptors were unable to detect the presence of interacting complexes between μ and κ receptors in heterologous cells (Jordan and Devi, 1999). However, a study used antibodies to endogenous μ and κ receptors to detect the presence of μ and κ interacting complexes in spinal cord membranes from female but not male rats (Chakrabarti et al., 2010) (Table 1). This study reported that the levels of μ-κ interacting complexes were dependent upon the stage of the oestrous cycle (Chakrabarti et al., 2010). These results suggest that the detection of μ-κ interacting complexes may be susceptible to the detergent conditions used in co-immunoprecipitation studies or may depend on the tissue used. In addition, it is possible that in cells/tissues expressing μ and κ receptors the μ-κ interacting complexes may be inducible only under certain physiological conditions or may be under the regulation of sex hormones. Further support for the probable formation of μ-κ heteromers came from BRET assays showing that the two receptors are close enough to directly interact in live cells (Wang et al., 2005a). Thus it appears that μ and κ receptors can form physiologically relevant heteromers and studies to explore the functional role of this heteromer in biological systems are needed.

Examination of the properties of the μ-κ heteromer using radioligand binding and [³⁵S]GTPγS assays revealed that the binding affinity of μ receptor agonists, such as DAMGO and endomorphin-1, is lower in cells coexpressing μ and κ receptors compared with cells expressing only μ receptors. In the case of the κ receptor agonists, U69593 and U50488H, no differences were observed between cells expressing μ-κ heteromers or κ receptors (Wang et al., 2005a) (Table 1). These findings indicate that heteromerization with κ receptors alters the binding properties of μ receptors.

A few studies have focused on identifying μ-κ heteromer-selective ligands (Chakrabarti et al., 2010; Yekkirala et al., 2011). Although strong evidence for the role of dynorphin 1–17 at μ-κ heteromers is not available, it has been reported that N-naphthoyl-β-naltrexamine (NNTA) selectively activates μ-κ heteromers (Table 4), and it is 50 times more potent than morphine, as an antinociceptive agent (Yekkirala et al., 2011) (Table 1). Moreover, intrathecal administration of NNTA leads to greater antinociceptive effect (∼two orders of magnitude) than intracerebroventricular administration, an effect not seen in μ receptor-knockout mice, which would suggest a higher degree of functional coupling between μ and κ receptors in the spinal cord (Yekkirala et al., 2011). In addition, the administration of NNTA does not lead to the development of physical dependence while antinociceptive tolerance to NNTA is low upon chronic intracerebroventricular administration and not observed upon chronic intrathecal administration (Yekkirala et al., 2011) (Table 1). This suggests that μ-κ heteromers play important roles in pain regulation and that they may be viable targets for the development of analgesics devoid of the unwanted side effects associated with chronic morphine administration.

μ and nociceptin/orphanin FQ receptor heteromers

Studies showing that ligands for the nociceptin/orphanin F/Q (NOP) receptor modulated the antinociceptive effects of morphine (King et al., 1998; Rizzi et al., 2000) led to investigations on whether μ receptors and NOP receptors formed heteromers. Co-immunoprecipitation studies showed that μ receptors formed interacting complexes with NOP receptors in heterologous cells coexpressing both receptors and in adult rat dorsal root ganglions (Pan et al., 2002; Wang et al., 2005b; Evans et al., 2010). In addition, studies using C-terminal deletion mutants suggested an involvement of the intracellular C-terminal region in the formation of these μ-NOP receptor complexes (Wang et al., 2005b) (Table 1).

Examination of the pharmacological properties of μ-NOP complexes showed an increase in the binding affinity of μ receptor agonists including DAMGO and fentanyl, compared with cells expressing only μ receptors (Pan et al., 2002) (Table 1). In addition, signalling assays showed that while the EC₅₀ for inhibition of adenylyl cyclase (AC) activity and for activation of ERK1/2 phosphorylation by DAMGO was significantly increased in cells coexpressing μ and NOP receptors, compared with cells expressing only μ receptors, the EC₅₀ for the NOP receptor agonist was not affected (Wang et al., 2005b). These findings taken together with the data from co-immunoprecipitation studies suggest that μ and NOP receptors form heteromers that selectively impair the potency of μ-induced signal transduction pathways. In addition, while pretreatment with the NOP receptor agonist caused desensitization of not only NOP- but also DAMGO-induced inhibition of AC activity, pretreatment with DAMGO did not affect desensitization of NOP receptor-mediated pathways (Wang et al., 2005b) (Table 1). These results suggest that μ-NOP heteromerization selectively causes cross-desensitization of μ receptor-mediated signal transduction.

A naltrexone derivative, IBN_txA, has been identified as a putative μ-NOP heteromer-selective ligand based on the high-affinity binding of radiolabelled IBN_txA in cells coexpressing μ1G receptors (a μ receptor with mutations in the sixth transmembrane region) and NOP receptors (Majumdar et al., 2011) (Table 4). This compound was 10 times more potent as an antinociceptive agent than morphine and did not display side effects, such as respiratory depression, physical dependence and appreciable constipation. Moreover, IBN_txA did not appear to be either rewarding or aversive in conditioned place preference studies (Majumdar et al., 2011) (Table 1). Taken together, these results suggest that targeting the μ-NOP heteromer could provide a major advance in the design and development of new highly potent opiate analgesics, without many side effects.

Heteromers between opioid and cannabinoid receptors

δ opioid and cannabinoid CB₁ receptor heteromers

Several lines of evidence have suggested interactions between δ-opioid receptors and CB₁ receptors. These included studies showing that (i) δ receptor agonists decrease CB₁ receptor signalling, (ii) δ receptor antagonists attenuate CB₁ receptor-mediated anxiolytic effects, (iii) CB₁ receptor levels and signalling increase in the substantia nigra of δ receptor-knockout mice and (iv) δ receptor activity increases in the caudate putamen of CB₁ receptor-knockout mice (Shapira et al., 1998; Berrendero and Maldonado, 2002; Berrendero et al., 2003; Uriguen et al., 2005). Direct interactions between δ and CB₁ receptors were suggested by BRET assays carried out in heterologous cells coexpressing luciferase-tagged CB₁ receptors and yellow fluorescent protein (YFP)-tagged δ receptors showing both receptors in close proximity in live cells (Rios et al., 2006). This was further supported by co-immunoprecipitation studies using epitope-tagged receptors and by immunofluorescence studies showing colocalization of δ receptors with CB₁ receptors in cortical neurons (Rozenfeld et al., 2012) (Table 2).

Table 2.

Heteromers involving opioid and cannabinoid/catecholamine receptors^a

Heteromer pair	In vitro heteromer properties (binding, signalling, trafficking)	In vivo effects of reagents targeting heteromers	References
δ-CB1b2006	Detection Colocalization, Co-IP, BRET Binding Remains to be determined. Signalling ↓ potency of CB1 receptor agonist. Heteromer signalling involves PLC-mediated recruitment of β-arrestin. Trafficking ↑ in CB1 receptor cell surface localization.	δ-CB1 heteromer-selective antibody Antibody detects ↑ in heteromer levels in cortex in neuropathic pain model (L5 spinal nerve ligation/transection). Antibody blocks CB1 receptor-mediated enhancement of δ receptor activity.	Rios et al., 2006; Bushlin et al., 2012; Rozenfeld et al., 2012
μ-CB1Rc2006	Detection Co-IP, BRET, FRET Binding Remains to be determined. Signalling ↓ in signalling with μ + CB1 receptor agonists (ERK1/2, Src, STAT3 phosphorylation). Trafficking CB1 receptor agonist or longer spacer bivalent ligand does not induce internalization, but μ agonist or mixture of both monovalent agonists induces heteromer internalization.	Bivalent ligand (μ agonist joined to CB1 receptor antagonist by a spacer arm) Antinociception (i.t. and i.c.v.) without development of tolerance.	Rios et al., 2006; Hojo et al., 2008; Le Naour et al., 2013
δ-α2A	Detection Colocalization, Co-IP, BRET Binding Remains to be determined. Signalling Remains to be determined. Trafficking Remains to be determined.	Combination of δ receptor agonist and α2A receptor agonist Synergy in antinociceptive effect (i.t.). ↑ in δ receptor-mediated neurite outgrowth.	Rios et al., 2004; Overland et al., 2009; Riedl et al., 2009
δ-β2	Detection Co-IP, BRET Binding No change. Signalling No change. Trafficking Internalized by δ receptor agonists and vice versa.		Jordan et al., 2001; Ramsay et al., 2002
κ-β2	Detection Co-IP, BRET Binding No change. Signalling No change in β2 adrenoceptor-mediated G-protein signalling but ↓ in β2 adrenoceptor-mediated ERK1/2 phosphorylation. Trafficking Suppression of β2 adrenoceptor agonist-mediated β2 adrenoceptor internalization.		Jordan et al., 2001; Ramsay et al., 2002
μ-α2A	Detection Colocalization, Co-IP, BRET, FRET Binding Remains to be determined. Signalling ↑ in μ receptor signalling in the presence of α2A adrenoceptor. ↓ in μ receptor signalling in the presence of α2A adrenoceptor ligands. Trafficking Controversial. One study showed no internalization of the heteromer but another study showed internalization of the heteromer and cross-desensitization by α2 adrenoceptor or μ agonist via p38 MAPK.		Jordan et al., 2003; Zhang and Limbird, 2004; Vilardaga et al., 2008; Tan et al., 2009

^aModified from (Gomes et al., 2013a).

^b,cThese heteromers are useful therapeutic targets, based on the known roles of one or both protomers. Therapeutic targets: ^bNeuropathic pain; ^cAntinociception.

Heteromerization with δ receptors modulates the subcellular localization of CB₁ receptors as the latter receptor exhibits an intracellular localization when expressed alone and when coexpressed with δ receptors, it was found on the cell surface (Rozenfeld et al., 2012). This differential localization requires the association of CB₁ receptors with the adaptor protein-2 whereas, in the absence of δ receptors, CB₁ receptors associate with adaptor protein-3 (Rozenfeld et al., 2012). Signalling assays showed that the δ receptor and CB₁ receptor heteromer (δ-CB₁ heteromer) exhibited signalling distinct from that of receptor homomers. For example, the signalling potency of a CB₁ receptor agonist was decreased in cells expressing the heteromer compared with the CB₁ receptor homomer and this decrease in potency was not seen following knockdown of δ receptors (Rozenfeld et al., 2012). Also, while δ or CB₁ receptor homomers signal via activation of Gα_i/o proteins, δ-CB₁ heteromer-mediated signalling involves PLC-mediated recruitment of β-arrestin 3 and the activation of signalling pathways that promote cell survival (Rozenfeld et al., 2012) (Table 2). Taken together, these findings show that δ-CB₁ heteromerization expands the signalling repertoire of individual receptors.

The observation that δ-CB₁ heteromer levels are altered under pathophysiological conditions is of great importance. In a rodent model of neuropathic pain where animals exhibit mechanical allodynia, δ-CB₁ heteromer levels (detected using a δ-CB₁ heteromer-selective monoclonal antibody) are significantly elevated in cortex, hypothalamus and midbrain (Bushlin et al., 2012). Moreover, in this model of neuropathic pain, the activation of G-protein mediated signalling by a CB₁ receptor-selective agonist was increased while that of a δ receptor-selective agonist was decreased; however, the latter was restored in the presence of a CB₁ receptor-selective agonist (Bushlin et al., 2012). These findings suggest that the pharmacological effects of δ and CB₁ receptors could be altered under conditions of neuropathic pain and that heteromer formation might be involved in these changes (Bushlin et al., 2012) (Table 2). Taken together, these studies suggest that the δ-CB₁ heteromer could be a target for the development of novel therapeutics to treat neuropathic pain.

μ opioid and CB₁ receptor heteromers

Possible formation of heteromers between μ receptors and CB₁ receptors was suggested by studies showing functional interactions between these receptors. For example, antinociceptive synergy was reported when using a combination of morphine and a CB₁ receptor agonist (Cichewicz, 2004). Moreover, μ receptor-knockout mice do not exhibit conditioned place preference for CB₁ receptor agonists (Ghozland et al., 2002) while the reinforcing effects of morphine as well as the severity of withdrawal symptoms from this drug are absent in CB₁ receptor-knockout mice (Ledent et al., 1999). Direct evidence for the formation of μ and CB₁ receptor heteromers (μ-CB₁ heteromer) came from proximity-based assays such as BRET and FRET that showed these two receptors were close enough to directly interact in live cells (Rios et al., 2006; Hojo et al., 2008). Furthermore, co-immunoprecipitation studies demonstrated the presence of interacting μ-CB₁ receptor complexes in heterologous cells coexpressing both receptors (Hojo et al., 2008).

Examination of the properties of the μ-CB₁ heteromer in heterologous and endogenous systems shows that while agonists to individual protomers can activate G-protein-mediated signalling and ERK phosphorylation, a combination of agonists to both protomers causes a decrease in signalling (Rios et al., 2006). Moreover, a study used receptors fused to chimeric G-proteins to show that the agonist for either protomer in the μ-CB₁ heteromer induces signalling by activating the same G-protein (Hojo et al., 2008). Heteromerization between μ and CB₁ receptors is of physiological relevance given that while agonists to either protomer cause an increase in neurite outgrowth in Neuro 2A cells, a combination of agonists to both receptors reduces neurite outgrowth by decreasing the phosphorylation of Src and STAT3 (Rios et al., 2006) (Table 2). Taken together, these studies show that μ-CB₁ heteromerization leads to modulation of individual protomer signalling and that this heteromer may have a physiological role.

In order to understand the physiological role of μ-CB₁ heteromers, bivalent ligands comprising a selective μ receptor agonist connected to a CB₁ receptor-selective antagonist/inverse agonist, via spacers of different lengths, have been generated (Le Naour et al., 2013). Among these, the ligand having a 20-atom spacer was found to bridge both receptors in the μ-CB₁ heteromer (Table 4) and this compound exhibited potent antinociceptive effects without development of antinociceptive tolerance (Le Naour et al., 2013) (Table 2). Because the development of antinociceptive tolerance is a key side effect of opiates and cannabinoids that limits their clinical use, the μ-CB₁ heteromer is a potential target for the development of analgesics with reduced side effects.

Heteromers involving opioid and catecholamine receptors

δ opioid and α_2A adrenoceptor heteromers

Functional interactions between δ and α_2A adrenoceptors suggested possible heteromerization between these two receptors. This included studies showing that attenuation of substance P-mediated antinociception was potentiated by a combination of α_2A adrenoceptor and δ receptor agonists (Fairbanks et al., 2000), and that synergistic antinociceptive interactions between these two receptors were observed in μ receptor-knockout but not in α_2A adrenoceptor-knockout mice (Stone et al., 1997; Fairbanks et al., 2002; Guo et al., 2003). Colocalization studies suggested that δ and α_2A adrenoceptors could form interacting complexes as these receptors were extensively colocalized to the same cells in the terminals of capsaicin-sensitive substance P-expressing primary afferent neurons (Riedl et al., 2009). In addition, BRET assays show that these two receptors were close enough to directly interact in live cells (Rios et al., 2004). Furthermore, co-immunoprecipitation studies using epitope-tagged receptors show that the two receptors form interacting complexes (Rios et al., 2004) (Table 2). Although a combination of colocalization, co-immunoprecipitation and BRET studies suggest the formation of δ-α_2A heteromers, not much is known about how interactions between these two receptors modulate the binding, signalling and trafficking properties of individual protomers. Such information is necessary in order to consider this receptor pair as a bona fide heteromer.

The physiological relevance of probable δ-α_2A heteromerization was examined by investigating agonist-mediated neurite outgrowth in Neuro 2A cells. These studies showed that coexpression of α_2A adrenoceptors could increase deltorphin II-(a δ receptor agonist) mediated neurite outgrowth in Neuro 2A cells. This suggests that α_2A adrenoceptors could allosterically modulate δ-receptor function (Rios et al., 2004). Another study examined the effects of δ receptor and α_2A adrenoceptor agonists on antinociception and found that a combination of agonists to the two receptors resulted in ∼30-fold increase in antinociceptive potency compared with administration of individual receptor agonists (Overland et al., 2009). Moreover, this increase in antinociceptive potency by a combination of δ receptor and α_2A adrenoceptor agonists was blocked by a PKC and not by a PKA inhibitor, whereas when each agonist was individually administered its potency was blocked by a PLC and not by a PKC inhibitor (Overland et al., 2009). The PKC isoform involved in these interactions between δ and α_2A adrenoceptor agonists has been recently identified as PKCε (Schuster et al., 2013). The antinociceptive synergy observed with a combination of δ and α_2A adrenoceptor agonists is thought to be due to the synergistic inhibition of the release of calcitonin gene-related peptide (CGRP) from the terminals of primary afferent neurons in the spinal cord (Overland et al., 2009) (Table 2). Together, these studies suggest a role for δ-α2A interacting receptor complexes in pain modulation but additional studies showing that disruption of the heteromer pair leads to changes in associating complexes and pain modulation would further support this point.

δ opioid and β₂ adrenoceptor heteromers

Studies showing that [Leu]enkephalin (a δ-receptor agonist) modulates β₂ adrenoceptor signalling and contraction in the heart suggested possible heteromerization between these two receptors (Pepe et al., 1997; Xiao et al., 1997). Co-immunoprecipitation studies showed that these receptors form interacting complexes at the cell surface of HEK-293 cells (Jordan et al., 2001). In addition, BRET studies showed that δ receptors and β₂ adrenoceptors were in close proximity and could directly interact in live cells. Moreover, the BRET signal was not changed in the absence or presence of agonists to both receptors leading to the suggestion that δ-β₂ heteromers are constitutively formed in cells coexpressing both receptors (Ramsay et al., 2002) (Table 2).

Examination of the ligand-binding properties of the δ-β₂ heteromer shows that heteromerization between these two receptors does not lead to alterations in the pharmacological properties of individual protomers (Jordan et al., 2001) (Table 2). In addition, the signalling properties of δ receptoragonists (inhibition of AC activity) are similar in cells expressing either δ receptors or the δ-β₂ heteromer (Jordan et al., 2001). Similarly, the signalling properties of β₂ adrenoceptor agonists (stimulation of AC activity) are similar in cells expressing either β₂ adrenoceptors or the δ-β₂ heteromer (Jordan et al., 2001). Moreover, the activation of MAPK induced by δ receptor or β₂ adrenoceptor agonists in δ-β₂ heteromer-expressing cells was similar to that observed in cells expressing individual receptors (Jordan et al., 2001) (Table 2). These results indicate that heteromerization of δ receptors with β₂ adrenoceptors does not significantly affect signalling by agonists to individual protomers (Table 2).

Examination of the trafficking properties of the δ-β₂ heteromer shows that treatment with either δ receptor or β₂ adrenoceptor agonists induces heteromer internalization (Jordan et al., 2001), while in cells expressing individual receptors, δ receptor agonists do not induce β₂ adrenoceptor endocytosis and β₂ adrenoceptor agonists do not induce δ-receptor endocytosis (Jordan et al., 2001). These findings suggest that heteromerization between δ receptors and β₂ adrenoceptors leads to alterations in the trafficking of individual receptors (Table 2).

Very little ‘direct evidence’ is available about the physiological role of δ-β₂ heteromers. Studies suggest that interactions between δ receptors and β₂ adrenoceptors may play a role in pathological conditions such as myocardial ischaemia (Huang et al., 2007). Thus further studies, using either heteromer-selective ligands/antibodies or agents that disrupt the heteromer, are needed to elucidate the role of the δ-β₂ heteromer during normal physiology and pathology.

κ-opioid receptor and β₂ adrenoceptor heteromers

Studies showing that κ receptors and β₂ adrenoceptors are present in the heart, and that a β₂ adrenoceptor agonist modulates radiolabelled ligand binding to κ receptors suggested possible heteromerization between these two receptors (Ventura et al., 1989; Tai et al., 1991). Co-immunoprecipitation studies carried out in HEK-293 cells coexpressing myc-tagged κ receptors and Flag-tagged β₂ adrenoceptors detected the presence of interacting complexes at the cell surface (Jordan et al., 2001). Furthermore, BRET assays show that the two receptors are close enough to directly interact in live cells (Ramsay et al., 2002) (Table 2). Taken together, these studies indicate that κ receptors and β₂ adrenoceptors can form heteromers in cells coexpressing both receptors.

Examination of the ligand-binding properties of κ-β₂ heteromers shows that there are no changes in binding affinity of individual receptor ligands, such as U-69593, nor-binaltorphimine and isoprenaline, when comparing with cells expressing either κ receptors or β₂ adrenoceptors (Jordan et al., 2001); this suggests that heteromerization between these two receptors does not lead to alterations in the pharmacological properties of individual receptor protomers (Table 2).

The signalling properties of the κ-β₂ heteromer were investigated using the AC and the MAPK phosphorylation assays. In the case of the AC assay, in cells expressing the κ-β₂ heteromer, there were no changes in the ability of κ receptor agonists to inhibit and of β₂ adrenoceptor agonists to stimulate enzyme activity compared with cells expressing individual receptors (Jordan et al., 2001). In the case of agonist-mediated MAPK activation, it was observed that in cells expressing the κ-β₂ heteromer, the κ receptor agonist but not the β₂ adrenoceptor agonist could induce ERK1/2 phosphorylation. These findings suggest that while heteromerization with κ receptors does not significantly affect the functional G-protein coupling properties of β₂ adrenoceptors, it may promote biased signalling at β₂ adrenoceptors by preserving G-protein-mediated signalling (i.e. AC activity) but impairing MAPK signalling (Jordan et al., 2001) (Table 2).

Examination of trafficking properties show that treatment with a β₂ adrenoceptor agonist, that induces β₂ adrenoceptor internalization in cells expressing only this receptor, does not induce β₂ adrenoceptor internalization in cells expressing the κ-β₂ heteromer (Jordan et al., 2001) (Table 2). This suggests that heteromerization between κ receptors and β₂ adrenoceptors modulates the trafficking properties of the β₂ adrenoceptors.

Very little is known about the physiological relevance of κ-β₂ heteromers. Thus further studies are needed to evaluate the role of this heteromer pair in normal physiology and pathology and particularly in cardiac pathology, given the presence of both receptors in the heart.

μ opioid and α_2A adrenoceptor heteromers

Studies showing functional interactions between μ receptors and α_2A adrenoceptors suggested that these two receptors could form heteromers. These included studies showing that a combination of μ receptor and α_2A adrenoceptor agonists resulted in antinociceptive synergy and that the potency of morphine-mediated antinociception is decreased in mice lacking α_2A adrenoceptors (Drasner and Fields, 1988; Ossipov et al., 1997; Stone et al., 1997). Several lines of evidence support heteromerization between μ receptors and α_2A adrenoceptors. For example, immunohistochemical analysis using Flag-tagged μ and hemagglutinin-tagged α_2A adrenoceptors demonstrate colocalization of both receptors not only at the plasma membrane but also within intracellular vesicles (Jordan et al., 2003). In addition, co-immunoprecipitation studies detect the presence of interacting μ-α_2A complexes in heterologous cells and in primary hippocampal neurons, and BRET and FRET assays show that the two receptors are close enough to form interacting complexes in live cells (Jordan et al., 2003; Zhang and Limbird, 2004; Vilardaga et al., 2008). Moreover, FRET analysis revealed that binding of an agonist to μ receptors suppressed the α_2A adrenoceptor agonist (norepinephrine)-induced FRET signal probably through a conformational change transmitted from the μ receptors to the α_2A adrenoceptors (Vilardaga et al., 2008) (Table 2). Interestingly, while treatment with agonists to individual receptors, such as morphine (μ receptor) or clonidine (α_2A adrenoceptor), leads to an apparent increase in the levels of μ-α_2A complexes, a combination of agonists to the two receptors leads to a decrease to below the basal levels (Jordan et al., 2003). This suggests that either co-occupancy of both protomers disrupts heteromer formation or makes the latter more susceptible to the effects of detergents used for cell lysis.

Examination of G-protein and MAPK activation in cells expressing the μ-α_2A heteromer shows that the morphine-mediated signalling is enhanced compared with cells expressing only μ receptors; however, this enhancement of signalling is not seen when a combination of morphine with clonidine (α₂ adrenoceptor agonist) is used (Jordan et al., 2003). Furthermore, the G-protein as well as MAPK activation by noradrenaline (α_2A adrenoceptor agonist) is decreased in the presence of morphine (Vilardaga et al., 2008). Similar findings were made with spinal cord neurons (Jordan et al., 2003), suggesting that the interactions between μ receptors and α_2A adrenoceptors also take place in endogenous systems and that the signalling of μ receptors can be modulated by α_2A adrenoceptors (Table 2). This is also supported by studies carried out in mouse dorsal root ganglion neurons which show that prolonged treatment with the μ receptor agonist, DAMGO, or the α₂ adrenoceptor agonist, clonidine, induced cross-desensitization between μ and α_2A receptor-mediated inhibition of voltage-gated Ca⁺² current and this was associated with the co-internalization of μ and α_2A adrenoceptors (Tan et al., 2009).

Although studies have shown that the analgesic potency of spinally administered morphine is decreased in mice lacking α_2A adrenoceptors (Stone et al., 1997), very little is known about the physiological relevance of μ-α_2A heteromers. Thus further studies, using either heteromer-selective ligands/antibodies or agents that disrupt the heteromer, are needed to elucidate the role of the μ-α_2A heteromer during normal physiology and pathology in particular during pain attenuation.

Other heteromers involving opioid receptors

δ opioid and chemokine receptor heteromers

The formation of heteromers between δ opioid receptors and the chemokine receptor CXCR4 (δ-CXCR4 heteromer) is suggested by FRET and co-immunoprecipitation studies carried out using heterologous cells and primary monocytes from healthy donors (Pello et al., 2008) (Table 3). The FRET signal and the level of interacting δ-CXCR4 complexes did not change in the presence or absence of receptor-selective agonists such as DPDPE (δ receptor agonist) or CXCL12 (CXCR4 agonist), suggesting that these heteromers are constitutively formed (Pello et al., 2008).

Table 3.

Heteromers involving opioid receptors and GPCRs other than cannabinoid or catecholamine receptors^a

Heteromer pair	In vitro heteromer properties (binding, signalling, trafficking)	In vivo effects of reagents targeting heteromers	References
δ-CXCR4	Detection Co-IP, FRET Binding No change. Signalling Inactivated by δ + CXCR4 agonists. No association with G-proteins in the presence of δ + CXCR4 agonists. Trafficking No change.		Pello et al., 2008
δ-SNSR-4	Detection BRET Binding Remains to be determined Signalling Preferential Gαq signalling and attenuation of Gαi signalling. Trafficking ↓ in δ receptor endocytosis by BAM22 in the presence of SNSR-4.		Breit et al., 2006
κ-APJ	Detection Colocalization, Co-IP, BRET Binding Remains to be determined. Signalling ↑ in heteromer-mediated PKC signalling. Trafficking Remains to be determined.		Li et al., 2012
μ-CCR5b2002	Detection Co-IP Binding Binding affinity for a ligand to one protomer not changed in the presence of ligand to partner protomer. Signalling ↓ in μ receptor -mediated G-protein activation by CCR5 receptor agonist and vice versa. Trafficking μ receptor internalization by μ and not by CCR5 receptor agonists and vice versa.	Bivalent ligand (Bivalent ligand 1) More potent inhibition of viral entry compared with naltrexone + maraviroc in antiviral activity assay.	Suzuki et al., 2002; Chen et al., 2004; Yuan et al., 2012; 2013
μ1D-GRPRc2002	Detection Co-IP Binding Not reported. Signalling μ receptor-mediated Ca+2 signalling only in cells expressing the heteromer. Trafficking ↑ in μ receptor -mediated GRPR internalization.	TAT-fusion protein (TAT-μ1DCT) TAT-μ1DCT disrupts μ1D-GRP receptor heteromers and blocks morphine-induced scratching without affecting analgesia.	Liu et al., 2011
μ-mGlu5d2002	Detection Co-IP Binding No change in binding affinity for μ receptor agonist. ↑ in binding affinity for mGlu5 receptor antagonist. Signalling No change in μ receptor agonist-mediated inhibition of adenylate cyclase activity. mGlu5 receptor antagonist ↓ μ receptor agonist-mediated phosphorylation and desensitization of μ receptors. Trafficking mGlu5 receptor antagonist ↓ μ receptor agonist-mediated internalization of μ receptors.	Bivalent ligand (MMG22) MMG22 antinociception (i.t. and i.c.v.) is equipotent to morphine in naïve mice, and more potent in LPS-treated mice with less tolerance and respiratory depression. MMG22 shows antinociception (i.t.) in CFA-induced inflammatory pain or bone cancer pain model.	Schroder et al., 2009; Akgun et al., 2013
μ-5-HT1A	Detection Colocalization, Co-IP, BRET Binding Remains to be determined. Signalling Transactivation of G-protein fused to 5-HT1A receptor by μ receptor agonist. Activation of ERK1/2 by μ receptor agonist is blocked by 5-HT1A receptor agonist pretreatment. Trafficking No co-internalization by protomer-selective agonists.		Daval et al., 1987; Pompeiano et al., 1992; Wang et al., 1998; Zhang et al., 2000; Kishimoto et al., 2001; Cussac et al., 2012
μ-NK1d2002	Detection Colocalization, Co-IP, BRET Binding ↑ in affinity for μ receptor agonist. No change in affinity for NK1 receptor agonist. Signalling Pre-incubation with μ agonist ↓ NK1 receptor-mediated ERK phosphorylation and vice versa. Trafficking Co-internalization by protomer-selective agonists.	Bivalent peptide Assay shows that the peptide exhibits μ agonist and NK1 receptor antagonist activity. Small molecule ligands Assay shows that the ligands exhibit μ agonist and NK1 receptor antagonist activity. Multifunctional μ/δ agonist/NK1 receptor antagonist compound (TY027) TY027 exhibits antinociception (i.c.v., i.t.) in naïve mice. TY027 exhibits antinociception (i.t., i.v.) against spinal nerve ligation-induced hyperalgesia. TY027 produced antinociception with low tolerance, dependence or rewarding effects and was not accompanied by opioid-related emesis or constipation.	Aicher et al., 2000a,b; Pfeiffer et al., 2003; Yamamoto et al., 2007; Vardanyan et al., 2011; Largent-Milnes et al., 2013
μ-sst2A	Detection Colocalization, Co-IP Binding ↓ in binding affinity for sst2 receptor agonists. No change in binding affinity for μ receptor agonist. Signalling ↑ inhibition of adenylate cyclase activity. No change in ERK1/2 activation. Pretreatment with the protomer agonist causes cross-desensitization. Trafficking sst2 receptor agonist induces heteromer internalization. μ agonist internalizes μ receptors but not sst2A receptors.		Pfeiffer et al., 2002

^aModified from (Gomes et al., 2013a).

^b,c,dThese heteromers are useful therapeutic targets, based on the known roles of one or both protomers. Therapeutic targets: ^bAntiviral activity; ^cMorphine-induced itch; ^dAntinociception.

Examination of intracellular signalling shows that although selective agonists (DPDPE or CXCL12) lead to Gα_i/o-protein activation in cells that coexpress both receptors, a combination of these two agonists inhibits receptor association with Gα_i/o protein (Pello et al., 2008). These observations suggest that while δ-CXCR4 heteromers are fully functional when activated by agonists to either protomer, a combination of agonists to both protomers inactivates the heteromer. Furthermore, CXCR12-induced phosphorylation of CXCR4 (or desensitization of CXCR4) was not altered by cotreatment with DPDPE (Pello et al., 2008) (Table 3), suggesting that the simultaneous activation of both protomers in δ-CXCR4 heteromers does not promote heterologous desensitization.

Very little is known about the physiological role of δ-CXCR4 heteromers. However, both δ receptors and CXCR4 are widely distributed in brain tissues and immune cells, and play key roles in inflammation processes and in pain sensation. As activation of both protomers in the δ-CXCR4 heteromer appears to result in a ‘silent’ receptor complex, further studies to evaluate the role of this heteromer in vivo, particularly under inflammatory conditions are needed.

δ opioid and sensory neuron-specific receptor (SNSR4) heteromers

The sensory neuron-specific receptor 4 is a GPCR with many names including SNSR3, SNSR4 and the official name of MRGPRX1 receptor; for brevity here it will be referred to as SNSR4. Heteromerization between δ receptors and SNSR4 was investigated based on studies showing that both receptors are present in dorsal root ganglia, and are activated by the bovine medulla adrenal peptide 22 (BAM22; a cleavage product of proenkephalin) although δ receptors mediate antinociceptive responses while SNSR4 mediates nociceptive responses (Lembo et al., 2002; Grazzini et al., 2004). BRET assays carried out in heterologous cells coexpressing δ and SNSR4 show that both receptors are in close proximity to one another and could directly interact in live cells (Breit et al., 2006) (Table 3). While δ receptor-selective agonists activate Gα_i/o-mediated signalling, SNSR4-selective agonists activate Gα_q-mediated signalling in cells expressing either δ receptors or SNSR4 or in cells coexpressing both receptors; this suggests that each receptor in the heteromeric complex acts as an independent signalling unit (Breit et al., 2006). Interestingly, naltrexone, an opioid receptor antagonist, can block BAM22-mediated Gα_q activation (Breit et al., 2006) suggesting transinhibition of SNSR4 signalling by δ receptors within the δ-SNSR4 heteromer. Furthermore, costimulation of both protomers in the δ-SNSR4 heteromer leads to preferential activation of Gα_q-mediated signalling (PLC activation) and inhibition of Gα_i/o-mediated signalling (Breit et al., 2006) (Table 3). This regulatory influence of SNSR4 on δ receptor signalling is not due to the PKC-mediated δ receptor desensitization (Breit et al., 2006). Similar observations were made with cultured dorsal root ganglia from rat embryos (Breit et al., 2006), indicating that these changes in δ receptor coupling and signalling because of heteromerization with SNSR4 also occur in vivo.

As both δ and SNSR4 contribute to the regulation of pain sensation (antinociception and nociception, respectively), further investigation on the role of the δ-SNSR4 heteromer in normal physiology and pathology is needed.

κ opioid and apelin receptor (APJ) heteromers

Studies have suggested the formation of heteromers between κ receptors and APJ receptors. Immunocytochemical studies in heterologous cells coexpressing both receptors show that both are colocalized predominantly at the plasma membrane (Li et al., 2012). Co-immunoprecipitation studies show that both receptors form interacting complexes and BRET assays show that they are in close enough proximity to directly interact in live cells (Li et al., 2012). Interestingly, treatment with receptor-specific agonists such as dynorphin A1–13 (for κ receptors) and apelin-13 (for APJ receptors) increased the BRET ratio, indicating that either the heteromerization between these two receptors was facilitated by receptor occupancy or that the latter induced conformational changes that decreased the distance between the epitope tags on individual protomers (i.e. between luciferase and YFP tags) (Li et al., 2012) (Table 3).

Signalling by the κ-APJ heteromer was examined in heterologous cells and in cells that endogenously express both receptors. These studies show that treatment with an agonist to either receptor induces a PKC-dependent ERK1/2 activation that is two- to threefold higher in cells coexpressing both receptors compared with cells expressing individual receptors (Li et al., 2012). In addition, heteromerization between κ and APJ receptors leads to an increase in PKC-mediated signalling and a decrease in PKA-mediated signalling compared with cells expressing individual receptors (Li et al., 2012) (Table 3). Taken together, these studies show that heteromerization between κ and APJ receptors leads to modulation of signalling by individual protomers.

The functional consequence of κ-APJ heteromerization was observed at the level of cell proliferation where treatment with either dynorphin A1–13 or apelin-13 significantly increased the proliferation of cells expressing the heteromer compared with cells expressing individual receptors (Li et al., 2012). Although not much is known about the role of the κ-APJ heteromer in vivo either in normal physiology or during pathology, the distribution of the dynorphin/κ receptor system and of the apelin/APJ system in the nuclei of the hypothalamus involved in regulation of arginine vasopressin release as well in the cardiovascular system (Sherman et al., 1986; Tsushima et al., 1993; Reaux et al., 2001), suggests a potential role for κ-APJ heteromers in cardiovascular regulation.

μ opioid and chemokine receptor CCR5 heteromers

Studies showing that μ-opioid receptors are present in immune cells and that morphine treatment increased the expression of the chemokine receptor CCR5 in lymphocytes led to investigations on heteromerization between these two receptors (Chuang et al., 1995; Miyagi et al., 2000). Co-immunoprecipitation studies show that CCR5 forms an interacting complex with μ receptors in cell lines that coexpress both receptors and that this is not modulated by treatment with receptor-selective ligands (Suzuki et al., 2002; Chen et al., 2004) (Table 3).

Examination of the pharmacological properties of the μ-CCR5 heteromer shows that pretreatment with the μ receptor agonist, DAMGO, did not change the binding of radiolabelled CCL4 (MIP-1β) to CCR5. Similarly, CCL5, another CCR5 ligand, did not change the binding properties of radiolabelled ligands to μ receptors (Chen et al., 2004) (Table 3). Interestingly, pretreatment with either DAMGO or CCL5 reduced CCL5- or DAMGO-mediated [³⁵S]GTPγS binding respectively (Chen et al., 2004). These results indicate that pretreatment with agonist to one protomer in the heteromeric complex reduces the ability of the partner receptor to activate G-proteins.

Examination of the trafficking properties of μ-CCR5 heteromers shows that the μ agonist, DAMGO, induced internalization of μ receptors and not of CCR5 while the CCR5 agonist, CCL5, induced internalization of CCR5 and not of μ receptors (Chen et al., 2004) (Table 3) suggesting that agonists selective for one receptor do not affect internalization of the other receptor in μ-CCR5 heteromers. Interestingly, pretreatment with either DAMGO or CCL5 can enhance the phosphorylation of both receptors in the heteromer, suggesting heterologous desensitization or cross-desensitization (Chen et al., 2004). In this context, activation of PKCζ has been reported to be involved in the cross-desensitization between μ and CCL5 (Song et al., 2011) (Table 3). The cross-desensitization between μ receptors and CCR5 within the heteromeric complex may modulate the physiological effects of opioids and chemokines in pathological conditions such as HIV infection or opiate addiction (Table 3).

A bivalent ligand targeting the μ-CCR5 heteromer that comprises a μ-selective antagonist pharmacophore, naltrexone, tethered through a 21-atom spacer to the CCR5-selective antagonist pharmacophore, maraviroc, has been developed (Yuan et al., 2012) (Table 4). This bivalent ligand is reported to be twice as potent as an inhibitor of viral entry, as a mixture of both antagonists in vitro (Yuan et al., 2013) (Table 3), suggesting a possible clinical usefulness of bivalent ligands targeting the μ-CCR5 heteromer against infection by HIV.

μ opioid 1D (μ1D) receptor and gastrin-releasing peptide (GRP) receptor heteromers

μ1D is a μ receptor isoform comprising exons 1–3 and 8–9 of the Oprm gene and μ1D receptors colocalize with GRP receptors (also known as bombesin BB₂ receptors) in the dorsal horn of the spinal cord (Liu et al., 2011). Co-immunoprecipitation studies with heterologous cells coexpressing both receptors or with spinal cord membranes show that μ1D and GRP receptors form interacting complexes (Liu et al., 2011) (Table 3).

Examination of the signalling properties of the μ1D-GRPR heteromer shows activation of the PLC-mediated Ca²⁺ signalling pathway by either morphine or GRP in cells coexpressing μ1D and GRP receptors (Liu et al., 2011). Moreover, morphine or GRP receptor-induced calcium spikes are blocked by the GRP receptor antagonist or by naloxone, indicating that morphine cross-activates GRP receptors through μ1D receptors (Liu et al., 2011) (Table 3). The μ1D-GRPR heteromer exhibits unique trafficking properties in that morphine treatment induces GRP receptor internalization in cells coexpressing both receptors but not in cells expressing only GRP receptors while a GRP receptor agonist did not induce μ1D receptor internalization in cells coexpressing both receptors (Liu et al., 2011). Taken together, these results suggest that heteromerization leads to modulation of protomer signalling and trafficking properties.

In order to elucidate the physiological roles of μ1D-GRPR heteromers, a membrane-permeable peptide consisting of TAT fused to μ1D_CT (TAT-μ1D_CT), that disrupts heteromer formation, has been developed. Intrathecal administration of TAT-μ1D_CT specifically blocks morphine-induced scratching without affecting morphine-induced analgesia (Liu et al., 2011) (Table 3). This suggests that the μ1D-GRPR heteromer may play a role in morphine-induced scratching. Moreover, the uncoupling of morphine-induced analgesia and morphine-induced scratching by the TAT-μ1D_CT peptide underscores the necessity for elucidating the function of individual μ receptor isoforms, which could be useful in the development of novel analgesics without side effects.

μ opioid and metabotropic glutamate mGlu₅ receptor heteromers

Co-immunoprecipitation studies show that μ and mGlu₅ receptors can form interacting complexes in HEK-293 cells coexpressing both receptors (Schroder et al., 2009). Interestingly, treatment with 2-methyl-6-(phenylethynyl) pyridine (MPEP), a mGlu₅ receptor antagonist, increases the levels of interacting complexes (Schroder et al., 2009); this suggests that occupancy of mGlu₅ receptors by MPEP affects the conformation of μ receptors which either facilitates the formation of μ-mGlu₅ heteromers or stabilizes the heteromer under the conditions used for receptor solubilization.

Examination of the pharmacological properties of μ-mGlu₅ heteromers shows that the binding affinity of the μ receptor agonist, DAMGO, is not changed when compared with cells expressing only μ receptors, while the binding affinity of the mGlu₅ receptor specific antagonist is increased when compared with cells expressing only mGlu₅ receptors (Schroder et al., 2009). In addition, while the presence of mGlu₅ receptors does not affect binding and signalling by μ receptors, occupancy of mGlu₅ receptors with the inhibitor MPEP causes a decrease in DAMGO-mediated phosphorylation, internalization and desensitization of μ receptors (Schroder et al., 2009). These findings suggest that a change in the conformation of mGlu₅ receptors by MPEP might allosterically regulate μ receptor function.

Given the wide expression of μ receptors and mGlu₅ receptors in the CNS and their role in regulation of pain transmission, opioid analgesia, dependence and withdrawal, ligands targeting this heteromer pair could play a role in pain regulation. In this context, a bivalent ligand targeting the μ-mGlu₅ heteromer has been developed. This ligand, MMG22, comprises a μ receptor agonist pharmacophore, oxymorphone, and a mGlu₅ receptor antagonist pharmacophore, m-methoxy-MPEP linked via a 22-atom spacer arm (Table 4). MMG22 exhibits antinociception similar to morphine in naïve mice (Akgun et al., 2013). However, it exhibits 4000 times more potent antinociception and less tolerance and respiratory depression compared with morphine in LPS-treated mice, an inflammatory pain model (Akgun et al., 2013). Taken together, these results suggest that MMG22 may be useful as a pharmacological tool to investigate μ-mGlu₅ heteromers in vivo, and in the development of novel drugs to treat inflammatory pain (Table 3).

μ opioid and 5-HT_1A receptor heteromers

Heteromerization between μ receptors and 5-HT_1A receptors was examined based on studies showing that (i) acute treatment with morphine increased 5-HT synthesis in different brain regions (Sastre-Coll et al., 2002), (ii) the inhibition of 5-HT synthesis by 5-HT_1A receptor agonists was enhanced in morphine-dependent animals (Sastre-Coll et al., 2002), (iii) chronic administration of a 5-HT_1A receptor agonist to the dorsal raphe nucleus delayed the development of tolerance to morphine (Nayebi and Charkhpour, 2006), (iv) chronic morphine administration increased 5-HT_1A receptor activity in the medial prefrontal cortex, and decreased 5-HT1A receptor activity in the dorsal raphe nucleus (Lutz et al., 2011) and (v) μ and 5-HT_1A receptors colocalize in discrete brain regions (Daval et al., 1987; Pompeiano et al., 1992; Wang et al., 1998; Zhang et al., 2000; Kishimoto et al., 2001). Co-immunoprecipitation and proximity-based assays carried out in cells coexpressing both receptors suggest that μ and 5-HT_1A receptors could form heteromers (μ-5-HT_1A) (Cussac et al., 2012) (Table 3).

Not much is known about the pharmacological properties of the μ-5-HT_1A heteromer. Examination of the signalling properties showed that the μ receptor agonist could activate a Gα_o protein that was covalently fused to 5-HT_1A receptors only in cells coexpressing both receptors (Cussac et al., 2012). In addition, phosphorylation of ERK1/2 induced following activation of μ receptors was blocked in the presence of the 5-HT_1A receptor agonist (Cussac et al., 2012). Examination of the trafficking properties of this heteromer shows that treatment with the agonist to one protomer did not induce internalization of the partner protomer (Cussac et al., 2012). Although these studies indicate that μ-5-HT_1A heteromerization may modulate the signalling properties of the μ receptor, further studies are needed to not only characterize the pharmacological properties of the μ-5-HT_1A heteromer but also its role during normal physiology and pathology.

μ opioid and substance P receptor heteromers

Electron microscopy studies demonstrating colocalization of μ receptors with the substance P NK₁ receptor, in the dendrites of the dorsal horn, together with co-immunoprecipitation and proximity-based assays showing that μ and NK₁ receptors form interacting complexes provide evidence for μ-NK₁ heteromerization (Aicher et al., 2000a,b; Pfeiffer et al., 2003) (Table 3).

Examination of the pharmacological properties of the μ-NK₁ heteromer shows that the μ receptor agonist, DAMGO, exhibits ∼threefold higher affinity compared with cells expressing only μ receptors, while substance P showed similar affinity as cells expressing only NK₁ receptors (Pfeiffer et al., 2003). Competition-binding assays showed that DAMGO or substance P did not compete with [³H]substance P or [³H]DAMGO, respectively, in μ-NK₁ heteromer expressing cells (Pfeiffer et al., 2003) (Table 3). However, both receptors in the μ-NK₁ heteromer can be cross-phosphorylated and co-internalized into the same endosomal compartment by protomer-selective agonists. This involves the recruitment of β-arrestin and formation of stable β-arrestin receptor complexes that are co-internalized (Pfeiffer et al., 2003) (Table 3). Taken together, these results indicate that although μ-NK₁ heteromerization does not affect the binding properties of the individual receptor protomers, it modulates their trafficking properties.

Activation of the μ-NK₁ heteromer by either DAMGO or substance P leads to rapid and transient MAPK activation (ERK1/2 phosphorylation) that is comparable with cells expressing individual receptors (Pfeiffer et al., 2003). However, pre-incubation with either DAMGO or substance P significantly attenuates either NK₁ or μ receptor-dependent ERK1/2 phosphorylation, respectively, suggesting that the μ-NK₁ heteromer undergoes homologous cross-desensitization (Pfeiffer et al., 2003). Moreover, the resensitization of μ receptor-mediated MAPK signalling is severely delayed in cells expressing μ-NK₁ heteromers as compared with cells expressing only μ receptors (Pfeiffer et al., 2003) (Table 3). These results indicate that the formation of heteromers between μ and NK₁ receptors influences the kinetics of signalling by individual protomers.

Both μ and NK₁ receptors play important roles in modulation of nociceptive responses. Therefore, targeting the μ-NK₁ heteromer could lead to the development of novel therapeutics to treat pain. In this context, attempts have been made to develop ligands that selectively target μ-NK₁ heteromers. These include bivalent peptides designed to possess the peptide sequences for both μ/δ receptor agonist and NK₁ receptor antagonist (Table 4). These bivalent peptides exhibited potent binding affinity and G-protein activation similar to that of either DPDPE or DAMGO alone (Yamamoto et al., 2007). Furthermore, the bivalent peptides exhibited agonistic activity for opioid receptors and antagonistic activity for NK₁ receptors in the guinea pig isolated ileum assay (Yamamoto et al., 2007). Further studies are needed to elucidate the effects of these peptides on pain regulation in vivo. Other ligands developed to selectively target the μ-NK₁ heteromer include small molecules that represent variations of combinations of structures of the opioid agonist fentanyl and the NK₁ receptor antagonist pharmacophore L732138 (Vardanyan et al., 2011); however, very little is known about the role of these small molecules in pain modulation. More recently, TY027, a multifunctional μ/δ receptor agonist/ NK₁ receptor antagonist compound, has been shown to have a preclinical profile of excellent antinociceptive efficacy, low abuse liability and no opioid-related emesis or constipation (Largent-Milnes et al., 2013). TY027 exhibited antinociceptive efficacy in both non-injured and spinal nerve-ligated animals (Largent-Milnes et al., 2013). In non-injured animals, the antinociceptive effect of TY027 was similar to that of morphine in the tail-flick test (Largent-Milnes et al., 2013). Moreover, repeated administration of TY027 did not lead to development of antinociceptive tolerance, dependence or reward (Largent-Milnes et al., 2013). Taken together, compounds targeting μ-NK₁ heteromers could be a promising therapeutic approach in treating patients who suffer from acute and chronic pain (Table 3).

μ opioid and somatostatin sst_2A receptor heteromers

The somatostatin sst_2A receptor and the μ receptor are closely related GPCRs that share ∼38% sequence homology (Pfeiffer et al., 2002). Heteromerization between these two GPCRs was suggested by immunohistochemical studies showing colocalization in the locus coeruleus (Pfeiffer et al., 2002). Furthermore, co-immunoprecipitation studies showed that μ and sst_2A receptors formed interacting complexes (Pfeiffer et al., 2002) (Table 3).

Pharmacological studies with μ-sst_2A heteromers showed that the binding affinities for the sst_2A receptor-selective agonist, L-779,976, were twofold lower than in sst_2A receptor homomers, while those for the μ receptor agonist, DAMGO, were similar to that of μ receptor homomers (Pfeiffer et al., 2002) (Table 3). Signalling studies showed that the ability of L-779,976 or DAMGO to inhibit AC activity was higher in cells coexpressing both receptors compared with cells expressing either sst_2A or μ receptors; however, no differences were observed for agonist-mediated ERK1/2 phosphorylation between cells expressing receptor heteromers and those expressing individual homomers (Pfeiffer et al., 2002). In addition, pretreatment with the agonist to one protomer attenuated signalling by the agonist to the partner protomer (Pfeiffer et al., 2002); this suggests cross-desensitization of signalling in μ-sst_2A heteromers. Furthermore, cross-phosphorylation of the receptors was also observed after treatment with receptor-selective agonists in cells coexpressing μ-sst_2A heteromers (Pfeiffer et al., 2002) (Table 3). Taken together, these results indicate that heteromerization between μ and sst_2A receptors leads to alterations in the signalling mediated via individual protomers.

Examination of the trafficking properties of the μ-sst_2A heteromer shows that while sst_2A receptor activation causes heteromer internalization, μ receptor activation induced only endocytosis of μ, but not of sst_2A receptors (Pfeiffer et al., 2002). These results suggest that μ-sst_2A heteromerization influences the trafficking properties of individual receptors (Table 3).

μ receptors and sst_2A receptors coexist and functionally interact in pain-processing pathways (Schulz et al., 1998), and some studies have described extensive cross-talk between opioid- and somatostatin-mediated analgesic responses (Betoin et al., 1994). However, further studies are needed to elucidate the in vivo role of μ-sst_2A heteromers in antinociception.

Concerns and limitations of techniques used to detect and characterize receptor heteromers

As described above, over the last decade, an increasing number of heteromers involving opioid receptors have been described. Research for the existence of most of these heteromers was brought about by evidence showing functional interactions between two receptor protomers. This gives rise to the question of whether a reported heteromer is truly a heteromeric complex where occupancy/activation of one protomer modifies binding at or signalling of the other protomer or whether the interactions between two receptors expressed in the same cells are the result of cross-talk because of activation of down-stream signalling cascades. A case in point involves δ opioid and dopamine D₁ receptors where studies show functional interactions between these two receptors (George and Kertesz, 1987; Daunais and McGinty, 1994; Unterwald and Cuntapay, 2000; Ito et al., 2006). Immunoelectron microscopic studies showing colocalization of both these receptors in the cytosol as well as in the plasma membrane of neurons in the striatum (Ambrose et al., 2006) is not sufficient to establish that the two receptors heteromerize as it does not demonstrate that the receptors are directly associated with each other. Therefore, further evidence for the formation of heteromers between δ and dopamine D₁ receptors, such as co-immunoprecipitation studies and proximity-based assays, are needed.

Most of the heteromers pairs described in this review were determined through the use of epitope-tagged receptors and a combination of co-immunoprecipitation studies and proximity-based assays such as BRET or FRET. It is to be noted that there are limitations to these techniques. For example, in the case of co-immunoprecipitation studies, the inherent hydrophobic nature of GPCRs could lead to the artifactual formation of interacting complexes. Thus when carrying out these studies, appropriate controls are needed such as (i) the use of tissue from animals lacking one of the receptors being investigated, (ii) using cells that express one of the two receptors, (iii) mixing cells that individually express each receptor prior to solubilization and immunoprecipitation, (iv) using different detergents for solubilization and (v) cross-linking cell surface proteins prior to solubilization and immunoprecipitation. Moreover, co-immunoprecipitation studies show that two receptors form interacting complexes but do not provide evidence for direct association between two receptors. Proximity-based assays, such as BRET or FRET, where a positive signal with/without ligand treatment would indicate that the two receptors are less than 100Å apart and therefore are directly associated with each other, have recently been used to provide support for direct association. However, a positive signal in these proximity-based assays could also be obtained because of overexpression of the receptors or because receptor occupancy leads to a movement of the acceptor and donor tags towards each other. Thus while carrying out proximity-based assays using epitope-tagged receptors, care should be taken to express both receptors at physiological levels and/or express different ratios of both receptors.

The limitations in the techniques used to ascertain whether two receptors form heteromers requires that a set of rules be applied to determine whether two receptors heteromerize. For in vitro studies, this would include the requirement that (i) the two receptors are present not only in the same cells but also in the same subcellular compartment; (ii) interacting receptor complexes can be isolated from cells that coexpress both receptors (but not from cells expressing only one receptor) using antibodies selective to the epitope tags on the receptors (and appropriate controls); (iii) proximity-based assays demonstrate that the donor or acceptor protein-tagged receptors exhibit energy transfer only in cells coexpressing receptors that are shown to interact by immunoprecipitation but not with receptors that have been shown not to interact; (iv) the heteromer exhibits a unique ‘signalling profile’ that is pharmacological, signalling or trafficking properties and that (v) the unique properties of the heteromer are blocked by agents (antibodies, ligands, peptides) that selectively recognize the heteromer. In a recent review, a set of criteria has been proposed to establish receptor heteromers in native tissue (Ferré et al., 2009). These include that: (i) the two receptors should localize not only in the same cells in a tissue but also in the same subcellular compartment; (ii) interacting receptor complexes should be immuno-isolated from wild-type animal tissue but not from animals lacking one of the receptors using receptor-selective antibodies; (iii) the signalling profile described for the heteromer pair in heterologous cells should match with that observed with native tissue; (iv) selective probes, such as antibodies or labelled ligands, should detect the presence of the heteromer in tissue sections; (iv) disruption of the heteromer through the use of agents that disrupt the heteromeric interactions should lead to reduced association and accompanying alterations of the signalling profile of the heteromer in native tissues.

A point to be kept in mind is that varying the ratio of individual receptor protomers would affect the level of receptor heteromers. Studies using proximity-based assays, such as BRET, have examined this point in detail and found that increasing the concentration of the YFP-tagged protomer while keeping the concentration of the luciferase-tagged protomer constant leads to a saturation of the detected BRET signal (this is an hyperbolic curve). These studies found that receptor heteromerization occurs at relatively low expression levels, leads to saturation, and that for optimal receptor heteromerization the two individual protomers have to be present at relatively equal levels (Canals et al., 2003; Terrilon et al., 2003; Breit et al., 2004). In our investigations of δ-μ receptor heteromerization, we found that increasing amounts of μ receptors into cells stably expressing δ receptors led to an initial increase followed by a decrease in δ-μ receptor interacting complexes as examined by co-immunoprecipitation (Rozenfeld and Devi, 2007). Similarly, increasing the levels of δ receptors (compared with μ) led to an initial increase and then a decrease in heteromer levels as examined by ligand-binding studies (Gomes et al., 2011). These results indicate that GPCR heteromerization is expression-dependent and that the two receptors need to be expressed at approximately equal levels for optimal heteromeric association.

The majority of the above criteria are met by only a handful of heteromers; this is due to the lack of tools available to probe the presence or regulation of heteromers in native tissue. The recent development of reagents, such as heteromer-selective antibodies (Gupta et al., 2010; Berg et al., 2012; Bushlin et al., 2012), TAT peptides that disrupt heteromers (He et al., 2011; Liu et al., 2011) or ligands targeting the heteromer (Bhushan et al., 2004; Daniels et al., 2005; Majumdar et al., 2011; Yekkirala et al., 2011; Yuan et al., 2012; Akgun et al., 2013; Gomes et al., 2013b; Le Naour et al., 2013), show promise that such studies are likely to move the field forward towards elucidating a role for GPCR heteromers in normal cell function and in pathology.

Conclusion

In the last decade, the formation of heteromers between opioid receptors and various GPCRs has been demonstrated and their unique intracellular signalling elucidated. Most studies used analytical techniques such as co-immunoprecipitation, colocalization, BRET and FRET assays in heterologous cells coexpressing epitope-tagged receptors to demonstrate the presence of receptor heteromers. Furthermore, recent GPCR crystal structural analysis provides support for the possibility that some GPCRs can form heteromers. Although these valuable techniques allow us to view the heteromers in vitro, when considering the physiological roles of heteromers in vivo, additional analytical tools are needed. As shown in this review, such in vivo analytical tools including selective ligands, antibodies and TAT-fusion proteins targeting selected opioid receptor heteromers are being developed and are providing clues to the physiological roles of these heteromers. For example, heteromer-selective antibodies help us ascertain the distribution of the heteromer in endogenous tissues and to detect changes in protein heteromer levels under specific pathological/therapeutical conditions such as neuropathic pain state or chronic drug treatment. Of great interest is the fact that many of the heteromer-selective ligands described in this review exhibited more or equipotent antinociceptive effects when compared with morphine but with lesser side effects such as development of tolerance or dependence after chronic administration. These findings strongly support the idea of using heteromers involving opioid receptors as targets for the development of novel therapeutic agents to treat pain.

Glossary

6′-GNTI

6′-guanidinonaltrindole

AC

adenylyl cyclase

APJ

apelin receptor

DAMGO

[D-Ala², N-MePhe⁴, Gly-ol]-enkephalin

DPDPE

D-penicillamine(2,5)-enkephalin

GRP

gastrin-releasing peptide

MPEP

2-methyl-6-(phenylethynyl) pyridine

NNTA

N-naphthoyl-β-naltrexamine

RTP4

receptor transport protein 4

SNSR4

sensory neuron-specific receptor 4

YFP

yellow fluorescent protein

Conflict of interest

The authors declare no conflict of interest.